Hi-Mini : A program for identification of high-order interaction

Hi-Mini allows an efficient high-order interaction analysis in genome-wide association studies using multifactor dimensionality reduction. It is a program based on character user interface (CUI), hence it does not provide graphical user interface. In order to use Hi-Mini, an understanding on command-line interface is required.


Hi-Mini supports multiple operating systems, including Windows, various Linux distributions and Mac operating system. Downloading Hi-Mini is possible via below links.

Input file

Hi-Mini can load various types of SNV dataset file, including PED file, Binary PED file and transposed PED file. For the details of above file formats, please refer PLINK website. The input option should be --bed for Binary PED file, --ped for PED file, and --tped for transposed PED file, respectively. Note that the phenotype field of input file should encoded in 1 for control and 2 for case.


himini --bed sample.bed : Assigns sample.bed as an input
himini --bed sample.bed --hmdr 8 --mdr --order 2 --top 1000 : Search up to 8th-order interaction using Hi-Mini, based on top 1000 results from second-order exhaustive search
himini --bed sample.bed --hmdr 8 --mdr --order 2 --top 1000 --hmdrprior top1000_order2.txt : Search up to 8th-order interaction using Hi-Mini, based on top 1000 results from pre-computed second-order exhaustive search

In case of third example, it is required to analyze second-order exhaustive MDR using cuGWAM or other software. The prior file of third example should consists of N lines for top N search, and each line should consists of P+1 columns. In each line, the first P elements are marker IDs in the dataset, and the last element should be balanced accuaracy of that interaction pair. Note that this P should equivalent to the parameter assigned in --order. For detailed instructions on cuGWAM, refer this page.